[Visualizing linear relationships] (http://seaborn.pydata.org/tutorial/regression.html#regression-tutorial)

Python实践：seaborn的散点图矩阵（Pairs Plots）可视化数据

image.png image.png image.png

使用PairGrid类的真正好处在于我们想要创建自定义函数来将不同的信息映射到图上。例如，我可能想要将两个变量之间的Pearson相关系数添加到散点图中。为此，我会编写一个函数，它接受两个数组、计算统计量，然后在图上绘制它。下面的代码显示了这是如何完成的（归功于这个Stack Overflow答案）：

# Function to calculate correlation coefficient between two arrays
def corr(x, y, **kwargs):
    # Calculate the value
    coef = np.corrcoef(x, y)[0][1]
    # Make the label
    label = r'$\rho$ = ' + str(round(coef, 2))
    # Add the label to the plot
    ax = plt.gca()
    ax.annotate(label, xy = (0.2, 0.95), size = 20, xycoords = ax.transAxes)
# Create a pair grid instance
grid = sns.PairGrid(data= df[df['year'] == 2007],
                    vars = ['life_exp', 'log_pop', 'log_gdp_per_cap'], size = 4)

# Map the plots to the locations
grid = grid.map_upper(plt.scatter, color = 'darkred')
grid = grid.map_upper(corr)
grid = grid.map_lower(sns.kdeplot, cmap = 'Reds')
grid = grid.map_diag(plt.hist, bins = 10, edgecolor =  'k', color = 'darkred');

---

seaborn直方图、散点图与回归分析图的绘制

image.png

---

seaborn pairgrid: using kdeplot with 2 hues

image.png image.png image.png

from scipy import stats
import seaborn as sns
import matplotlib

def corrfunc(x, y, **kws):
    r, _ = stats.pearsonr(x, y)
    ax = plt.gca()
    # count how many annotations are already present
    n = len([c for c in ax.get_children() if 
                  isinstance(c, matplotlib.text.Annotation)])
    pos = (.1, .9 - .1*n)
    # or make positions for every label by hand
    pos = (.1, .9) if kws['label'] == 'Yes' else (.1,.8)
 ## 需特别注意有没有label
    ax.annotate("{}: r = {:.2f}".format(kws['label'],r),
                xy=pos, xycoords=ax.transAxes)

tips = sns.load_dataset("tips")
g = sns.PairGrid(data = tips, vars = ['tip', 'total_bill'], hue="smoker", size=4)
g.map_upper(plt.scatter, s=10)
g.map_diag(sns.distplot, kde=False)
g.map_lower(sns.kdeplot, cmap="Blues_d")
g.map_lower(corrfunc)
g.add_legend()

官网 Building structured multi-plot grids